CN110418367A - A kind of 5G forward pass mixture of networks edge cache low time delay method - Google Patents
A kind of 5G forward pass mixture of networks edge cache low time delay method Download PDFInfo
- Publication number
- CN110418367A CN110418367A CN201910515104.9A CN201910515104A CN110418367A CN 110418367 A CN110418367 A CN 110418367A CN 201910515104 A CN201910515104 A CN 201910515104A CN 110418367 A CN110418367 A CN 110418367A
- Authority
- CN
- China
- Prior art keywords
- web content
- mobile subscriber
- request
- cache
- content
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W24/00—Supervisory, monitoring or testing arrangements
- H04W24/06—Testing, supervising or monitoring using simulated traffic
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W28/00—Network traffic management; Network resource management
- H04W28/02—Traffic management, e.g. flow control or congestion control
- H04W28/10—Flow control between communication endpoints
- H04W28/14—Flow control between communication endpoints using intermediate storage
Abstract
The invention discloses a kind of 5G forward pass mixture of networks edge cache low time delay methods, first construct Core server, 5G forward pass network edge buffer network scene between cell base station and user, requested Web content title and corresponding request number of times are extracted from the request that mobile subscriber issues Web content again, safeguard into information matrix and clustering processing, then the caching value of each Web content in each class cluster is calculated, and base station and hybrid cache strategy of each mobile subscriber when storing Web content are formulated with this, and establish EMD-ARIMA prediction model and cache replacement policy, last mobile subscriber obtains the Web content needed, the caching redundancy between user is reduced in this way, improve network resource utilization, it reduces the service time delay of 5G forward pass network and alleviates the link stress of network.
Description
Technical field
The invention belongs to wireless communication technology fields, more specifically, it is slow to be related to a kind of 5G forward pass mixture of networks edge
Deposit low time delay method.
Background technique
In recent years, with emerging technologies such as the high speed development of mobile Internet, various emerging communication services and Internet of Things
Appearance, pushed the development of 5G mobile communications network.In addition, mobile intelligent terminal is universal so that mobile data flow is held
It is continuous to increase.Estimate there will be about 50,000,000,000 shiftings with network connecting function in the year two thousand twenty global range according to industrial quarters
Dynamic terminal equipment access network, and need to be promoted 1000 times in the capacity of following 5th Generation Mobile Communication System.
Terminal magnanimity connection, flow surge all to existing communication system message transmission rate, spectrum efficiency and
Network capacity etc. causes huge pressure, and the load pressure of server also increases severely therewith, causes server timely
Each user request is responded.Secondly, the bandwidth of 5G forward pass network is limited, hair is easy in the peak traffic period
Raw network congestion causes Web content transmission to be obstructed, so that user experience is impaired.
Effective solution method is that caching is disposed on the cell base station or mobile terminal device for having caching capabilities, will be used
The content storage that family can request that in the buffer, obtains by this method to alleviate the pressure of 5G forward pass network link and reduce user
Take the time delay of content.The cache policy of mainstream is popularity caching at present, i.e., Web content is carried out descending row according to popularity
Then column cache the forward Web content of popularity on base station or mobile terminal device.However this cache way is not
In view of the concrete composition classification (such as staff or student etc.) of each community user, they are at the demand to Web content
Different, and they are also not necessarily included in most popular Web content the request of Web content, so cache policy
The specific request of each class users should be comprehensively considered, cached on demand；In addition request of each user to heterogeneous networks content
Grade is also different, i.e., there are difference for the hobby of each user.If the request of mobile subscriber not to be included in the column of analysis,
And the only popularity of simple consideration Web content, then between different equipment can maximum probability cache in identical network
Hold, the problem of largely caching redundancy is caused, so that spatial cache cannot be utilized adequately.
On the other hand, the popularity of Web content has time variation, i.e. user can at any time not to the request of Web content
Disconnected variation, this also causes traditional passive cache policy to be difficult to obtain preferable caching effect.If being capable of accurately pre- survey grid
Variation of the network user to content requests, so that it may be led in advance according to the network content request data of prediction in non-traffic peak period
The Web content that dynamic caching future time user can request that, this can not only mitigate the 5G forward pass network of peak traffic phase
Link stress, and the Internet resources of peak absences can be made full use of.Therefore, how accurately to predict that user requests net
The variation of network content has become the critical issue for promoting active cache strategy performance.
Summary of the invention
It is an object of the invention to overcome the deficiencies of the prior art and provide a kind of 5G forward pass mixture of networks edge cache is low
Time delay method predicts each mobile subscriber by EMD-ARIMA prediction model to the request situation of Web content, to reduce
Caching redundancy between user improves network resource utilization, the service time delay for reducing 5G forward pass network and alleviation network
Link stress.
For achieving the above object, a kind of 5G forward pass mixture of networks edge cache low time delay method of the present invention, feature
It is, comprising the following steps:
(1), the 5G forward pass network edge buffer network scene between Core server, cell base station and user is constructed；
Cell base station obtains Web content by 5G forward pass network from Core server, then is with communication
All mobile subscribers in entire cell range provide service, and every mobile subscriber mentions again in radiation scope for other users
For service；
(2), the request that mobile subscriber issues Web content extracts the requested Web content name of the mobile subscriber
Title and corresponding request number of times；
Content name and corresponding request number of times are formed into information matrix again, and are stored in the database of cell base station
In；Wherein, every a line of information matrix represents a mobile subscriber to the request number of times of each Web content, and each column represent one
Item Web content title；
(3), the data in information matrix are pre-processed
Information matrix is numbered by row from 1 to N, then each column of information matrix are subjected to minimax normalization
Processing；
Wherein, x indicates request number of times of the mobile subscriber to some Web content, xminIndicate the smallest request time in the column
Number, xmaxIndicate maximum request number of times in the column；
(4), the data after pretreatment are clustered using density peaks clustering algorithm
(4.1), the distance between mobile subscriber is calculated
If each mobile subscriber makes requests M Web content, any two are calculated using Euclidean distance formula and are moved
Employ family usAnd utThe distance between dist (us,ut)；
Wherein, s, t ∈ [1, N], and s ≠ t,Indicate mobile subscriber usTo the request number of times of m-th of Web content；
(4.2), mobile subscriber u is calculatedsLocal density ρs；
ρs=∑ χ (dist (us,ut)-distcutoff)
Wherein, distcutoffIndicate truncation distance, function χ (x) meets:
(4.3), mobile subscriber u is calculatedsWith the higher user u of local densitytThe distance between δs；
As mobile subscriber usLocal density when being maximum value, then mobile subscriber usWith remaining mobile subscriber utBetween away from
From δsAre as follows:
(4.4), it clusters
For all mobile subscribers, with local density ρsFor horizontal axis, distance δsFor the longitudinal axis, drawn certainly in plane coordinates
Plan figure；
The point for being located at the decision diagram upper right corner is selected as cluster class center again, remaining each point is belonged to apart from it recently
The affiliated class cluster in cluster class center in, to clustering all points for L class, L={ l1,l2,…,lτ,…,lL, lτFor τ class
In mobile subscriber's quantity；
(5), the caching value of each Web content in each class cluster is calculated
The caching value of p-th of Web content in class cluster τ are as follows:
Wherein, fk,pFor the element value in information matrix, i.e. expression user ukTo the request number of times of Web content p；
After the completion of the caching value calculation of each Web content in each class cluster, by all caching value maintenances at
Size is the matrix Score of L*M；
(6), base station and hybrid cache strategy of each mobile subscriber when storing Web content are formulated
(6.1), node B cache strategy is formulated
Each of matrix Score element is subjected to descending arrangement, then by Web content corresponding to preceding B element
It stores into the spatial cache of base station, if Web content corresponding to some element duplicates, stores in order next
A Web content；
(6.2), mobile subscriber's cache policy is formulated
Each mobile subscriber caches the highest b Web content of respective request number of times in memory space；
(7), EMD-ARIMA prediction model is established
(7.1), using EMD decomposition algorithm by mobile subscriber's request data x ' decomposition in all kinds of clusters obtained in step (3)
For multiple intrinsic modal components imfi(t), i indicates i-th of intrinsic modal components；
(7.2)、
To each imfi(t) ARIMA prediction model is constructed respectively；
Calculate each imfi(t) auto-correlation coefficientAnd PARCOR coefficients
Wherein, γ is intrinsic modal components imfi(t) length, h are delay period number, αtFor imfi(t) sample value,For imfi(t) desired value；
Wherein,
Using delay period number as abscissa, auto-correlation coefficient is that ordinate draws autocorrelogram；It is cross with delay period number
Coordinate, PARCOR coefficients are that ordinate draws partial autocorrelation figure；
Autocorrelogram is recycled to examine mobile subscriber to the stationarity of network content request data, if data are unstable,
Difference processing then is carried out to data, until data are steady, the number of accumulative difference processing is denoted as d, and predicts mould as ARIMA
The parameter of type；
Select suitable ARIMA pre- according to the truncation of time series autocorrelogram and partial autocorrelation figure and hangover property
Model is surveyed, and ARIMA prediction model order is determined according to BIC criterion；
Wherein, n indicates each imfi(t) in component sample data number, c is constant, and y is unknown parameter number, and ε is
White noise sequence,For the variance of ε；
ARIMA prediction model is constructed according to the parameter of ARIMA prediction model and order, the ARIMA of construction is recycled to predict
Model is to imfi(t) component is predicted, and its predicted value is denoted as imfi(t)'；
(7.3), by all intrinsic modal components imfi(t) predicted value imfi(t) ' be added and used as mobile in all kinds of clusters
The predicted value of family request data；
(8), cache replacement policy is formulated
(8.1), request time of the mobile subscriber to Web content in all kinds of clusters of EMD-ARIMA prediction model prediction is utilized
Number, then calculate the caching value Score' of each request content in all kinds of clustersτ(p), it and in descending order arranges；
(8.2), not buffered Web content is judged whether there is, if there is not buffered Web content m', and
And its caching value Score'τ(p) it is higher than some and is stored in the Web content m's in base station or mobile subscriber's spatial cache
Scoreτ(p), then Web content m is then replaced with m' in advance in non-traffic peak period；
(9), mobile subscriber obtains Web content
(9.1), mobile subscriber issues some Web content and requests, and mobile subscriber equipment terminal first can be at itself
It searches whether to store the Web content in advance in spatial cache, if being stored with the Web content, is denoted as own cache,
And this content is directly extracted from own cache space, the time used ignores；Otherwise, user equipment terminal to service model
Other users in enclosing broadcast the request to Web content, enter step (9.2)；
(9.2), it after other mobile subscribers receive request, searches whether to be stored in this from own cache space respectively
Hold, if being stored with the content, is denoted as neighborhood caching, and forward the Web content to request user, while recording institute's used time
Between t1；Otherwise uncached to request user's reply, and enter step (9.3)；
(9.3), mobile subscriber sends to cell base station and requests, and after cell base station receives request, searching in node B cache is
It is no to be stored with requested Web content, if so, being then denoted as node B cache, and the content is submitted to request user, recorded simultaneously
Time t used2；Otherwise, the content is downloaded by 5G forward pass link in base station from Core server, is then forwarded to request and uses
Family, while recording time t used3。
Goal of the invention of the invention is achieved in that
A kind of 5G forward pass mixture of networks edge cache low time delay method of the present invention, first constructs Core server, cell base station
And the 5G forward pass network edge buffer network scene between user, then from the request that mobile subscriber issues Web content
Requested Web content title and corresponding request number of times are extracted, then maintenance is calculated at information matrix and clustering processing
The caching value of each Web content in each class cluster, and base station and each mobile subscriber are formulated in storage Web content with this
When hybrid cache strategy, and establish EMD-ARIMA prediction model and cache replacement policy, last mobile subscriber, which obtains, to be needed
The Web content wanted reduces the caching redundancy between user in this way, improves network resource utilization, reduces 5G forward pass network
It services time delay and alleviates the link stress of network.
Meanwhile a kind of 5G forward pass mixture of networks edge cache low time delay method of the present invention also has the advantages that
(1), the present invention has fully considered all types of user in cell to the specific request situation of heterogeneous networks content and each
From the request rank to Web content, the request of every network user can be met with maximum probability possibility, improves caching life
Caching redundancy between middle rate, reduction user；
(2), request of all types of user to Web content is relatively accurately predicted by establishing EMD-ARIMA prediction model
Situation guarantees to use so as to shift to an earlier date the Web content that active cache future time user can request that in non-traffic peak period
Family experience.
Figure of description
Fig. 1 is a kind of 5G forward pass mixture of networks edge cache low time delay method flow diagram of the present invention；
Fig. 2 is 5G forward pass network edge caching schematic diagram of a scenario；
Fig. 3 is the spatial distribution schematic diagram of mobile subscriber in the cell；
Fig. 4 is the auto-correlation coefficient figure of user's request data；
Fig. 5 is the PARCOR coefficients figure of user's request data；
Fig. 6 is the effect contrast figure of EMD-ARIMA prediction model and ARIMA prediction model；
When Fig. 7 is user's request of the mentioned hybrid cache strategy of the present invention, random cache strategy and popularity cache policy
Prolong the effect contrast figure changed with Zipf rule alpha parameter；
When Fig. 8 is user's request of the mentioned hybrid cache strategy of the present invention, random cache strategy and popularity cache policy
Prolong the effect contrast figure changed with cell range；
When Fig. 9 is user's request of the mentioned hybrid cache strategy of the present invention, random cache strategy and popularity cache policy
Prolong the effect contrast figure with spatial cache volume change.
Specific embodiment
A specific embodiment of the invention is described with reference to the accompanying drawing, so that those skilled in the art is more preferable
Ground understands the present invention.Requiring particular attention is that in the following description, when the detailed description of known function and design
When perhaps can desalinate main contents of the invention, these descriptions will be ignored herein.
Embodiment
Fig. 1 is a kind of 5G forward pass mixture of networks edge cache low time delay method flow diagram of the present invention.
In the present embodiment, as shown in Figure 1, a kind of 5G forward pass mixture of networks edge cache low time delay method of the present invention, packet
Include following steps:
5G forward pass network edge buffer network scene between S1, building Core server, cell base station and user；
In the present embodiment, as shown in Fig. 2, establishing caching scene for single subdistrict, include in edge cache network
1 cell base station, N number of mobile subscriber and M Web content, mobile subscriber's spatial distribution in the cell as shown in figure 3,
All-network content size is identical.
Cell base station obtains Web content by 5G forward pass network from Core server, then is with communication
All mobile subscribers in entire cell range provide service, and every mobile subscriber mentions again in radiation scope for other users
For service.
The request that S2, mobile subscriber issue Web content extracts the requested Web content title of the mobile subscriber
And corresponding request number of times；
Content name and corresponding request number of times are formed into a N*M size information matrix again, and are stored in cell base
In the database stood；Wherein, every a line of information matrix represents a mobile subscriber to the request number of times of each Web content, often
One column represent a Web content title；
S3, the data in information matrix are pre-processed
Information matrix is numbered by row from 1 to N, distinguishing different users, then by each column of information matrix
Minimax normalized is carried out, to simplify subsequent processes；
The process of normalized are as follows:
Wherein, x indicates request number of times of the mobile subscriber to some Web content, xminIndicate the smallest request time in the column
Number, xmaxIndicate maximum request number of times in the column；
S4, the data after pretreatment are clustered using density peaks clustering algorithm
The distance between S4.1, calculating mobile subscriber
If each mobile subscriber makes requests M Web content, any two are calculated using Euclidean distance formula and are moved
Employ family usAnd utThe distance between dist (us,ut)；
Wherein, s, t ∈ [1, N], and s ≠ t,Indicate mobile subscriber usTo the request number of times of m-th of Web content；
S4.2, mobile subscriber u is calculatedsLocal density ρs；
ρs=∑ χ (dist (us,ut)-distcutoff)
Wherein, distcutoffIndicate truncation distance, function χ (x) meets:
ρsCalculation formula, which is meant that, to be found and mobile subscriber usThe distance between be less than truncation distance distcutoffShifting
The number at family is employed, and as usLocal density；
S4.3, mobile subscriber u is calculatedsWith the higher user u of local densitytThe distance between δs；
As mobile subscriber usLocal density when being maximum value, then mobile subscriber usWith remaining mobile subscriber utBetween away from
From δsAre as follows:
S4.4, cluster
For all mobile subscribers, with local density ρsFor horizontal axis, distance δsFor the longitudinal axis, drawn certainly in plane coordinates
Plan figure；
The point for being located at the decision diagram upper right corner is selected as cluster class center again, remaining each point is belonged to apart from it recently
The affiliated class cluster in cluster class center in, to clustering all points for L class, L={ l1,l2,…,lτ,…,lL, lτFor τ class
In mobile subscriber's quantity；
S5, the caching value for calculating each Web content in each class cluster
The caching value of p-th of Web content in class cluster τ are as follows:
Wherein, fk,pFor the element value in information matrix, i.e. expression user ukTo the request number of times of Web content p；
After the completion of the caching value calculation of each Web content in each class cluster, by all caching value maintenances at
Size is the matrix Score of L*M；
S6, base station and hybrid cache strategy of each mobile subscriber when storing Web content are formulated
S6.1, node B cache strategy is formulated
Each of matrix Score element is subjected to descending arrangement, then by Web content corresponding to preceding B element
It stores into the spatial cache of base station, if Web content corresponding to some element duplicates, stores in order next
A Web content, for example, the 4th and the 5th corresponding Web content of element duplicates, then it is corresponding in the 5th element
The 6th corresponding Web content of element is stored in spatial cache；
S6.2, mobile subscriber's cache policy is formulated
Each mobile subscriber caches the highest b Web content of respective request number of times in memory space；
S7, EMD-ARIMA prediction model is established
S7.1, mobile subscriber's request data x ' in all kinds of clusters obtained in step S3 is decomposed into using EMD decomposition algorithm
Multiple intrinsic modal components imfi(t), i indicates i-th of intrinsic modal components；
Intrinsic mode function meets following two condition:
(1), in entire time range, the extreme point number of each IMF component is necessarily equal to zero crossing number, or
At most difference 1；
(2), the mean value for the envelope that at any time, the local maximum of IMF component and local minimum are formed is equal to 0.
The detailed process of EMD decomposition algorithm is described in we below:
S7.1.1, the time series for changing mobile subscriber request data x ' at random according to request time, can be denoted as x '
(t), all Local Extremums for finding out mobile subscriber request data x ' (t), are fitted all using cubic spline functions
Maximum point and minimum point obtain the coenvelope line x of x ' (t)max(t) and lower envelope line xmin(t)；
S7.1.2, average value of the envelope on each time point up and down is calculated, obtains Mean curve m (t):
S7.1.3, h (t)=x'(t is enabled)-m (t), if h (t) meets the condition of intrinsic mode function, h (t) is x '
(t) first IMF component, is denoted as imf1(t)；If h (t) is unsatisfactory for condition, x'(t is enabled)=h (t), return again to step
Rapid S7.1.1；
S7.1.4, first IMF component imf is subtracted from x ' (t)1(t), obtain removing the residual components r of radio-frequency component1
(t), r1(t)=x'(t)-imf1(t)；
S7.1.5, by r1(t) after handling according to step S7.1.1~S7.1.4 the method, it can be obtained second IMF points
Measure imf2(t), r2(t)=r1(t)-imf2(t)；
S7.1.6, then and so on, by n1After secondary interative computation, if discrepanceTend to dull or only has
One pole, then calculating process stops；
S7.2, to each imfi(t) ARIMA prediction model is constructed respectively；
Calculate each imfi(t) auto-correlation coefficientAnd PARCOR coefficients
Wherein, γ is intrinsic modal components imfi(t) length, h are delay period number, αtFor imfi(t) sample value,For imfi(t) desired value；
Wherein,
Using delay period number as abscissa, auto-correlation coefficient is that ordinate draws autocorrelogram；It is cross with delay period number
Coordinate, PARCOR coefficients are that ordinate draws partial autocorrelation figure；
Autocorrelogram is recycled to examine mobile subscriber to the stationarity of network content request data, if data are unstable,
Difference processing then is carried out to data, until data are steady, the number of accumulative difference processing is denoted as d, and predicts mould as ARIMA
The parameter of type；
Select suitable ARIMA pre- according to the truncation of time series autocorrelogram and partial autocorrelation figure and hangover property
Model is surveyed, and ARIMA prediction model order is determined according to BIC criterion；
Wherein, n indicates each imfi(t) in component sample data number, c is constant, and y is unknown parameter number, and ε is
White noise sequence,For the variance of ε；
ARIMA prediction model is constructed according to the parameter of ARIMA prediction model and order, the ARIMA of construction is recycled to predict
Model is to imfi(t) component is predicted, and its predicted value is denoted as imfi(t)'；
S7.3, by all intrinsic modal components imfi(t) predicted value imfi(t) ' be added and used as mobile in all kinds of clusters
The predicted value of family request data；
S8, cache replacement policy is formulated
S8.1, request time of the mobile subscriber to Web content in all kinds of clusters of EMD-ARIMA prediction model prediction is utilized
Number, then calculate the caching value Score' of each request content in all kinds of clustersτ(p), it and in descending order arranges；
S8.2, not buffered Web content is judged whether there is, if there is not buffered Web content m', and
And its caching value Score'τ(p) it is higher than some and is stored in the Web content m's in base station or mobile subscriber's spatial cache
Scoreτ(p), then Web content m is then replaced with m' in advance in non-traffic peak period；This can not only alleviate peak traffic
The link stress of the 5G forward pass network of phase, and the Internet resources of non-peak period can also be made full use of.In addition, actively delaying in advance
Deposit the online experience that can guarantee mobile subscriber with replacement policy.
S9, mobile subscriber obtain Web content
S9.1, mobile subscriber issue some Web content and request, and mobile subscriber equipment terminal first can be in the slow of itself
It deposits and searches whether to store the Web content in advance in space, if being stored with the Web content, be denoted as own cache, and
This content is directly extracted from own cache space, the time used ignores；Otherwise, user equipment terminal is to service range
Interior other users broadcast the request to Web content, enter step S9.2；
After S9.2, other mobile subscribers receive request, search whether to be stored in this from own cache space respectively
Hold, if being stored with the content, is denoted as neighborhood caching, and forward the Web content to request user, while recording institute's used time
Between t1；Otherwise uncached to request user's reply, and enter step S9.3；
S9.3, mobile subscriber send to cell base station and request, and after cell base station receives request, searching in node B cache is
It is no to be stored with requested Web content, if so, being then denoted as node B cache, and the content is submitted to request user, recorded simultaneously
Time t used2；Otherwise, the content is downloaded by 5G forward pass link in base station from Core server, is then forwarded to request and uses
Family, while recording time t used3。
Experiment simulation
In order to verify the performance of cache policy and replacement policy proposed by the invention, random cache strategy and stream are selected
Object is emulated row degree caching as a comparison.150 mobile subscriber's random scatters are shared in this emulation in the cell, it is small
Area's side length value between 50~500.User meets Zipf law to the request probability of certain Web content:
Wherein, k is the Web content that caching value comes kth position, and the value range of α is 0.7~1.6, fkIt is used to be mobile
Request probability of the family to preceding k-th of Web content.Node B cache space B value range is 5~14, and the caching of mobile subscriber is empty
Between b be 2.Request time t1For 10ms, request time t2For 20ms, request time t3For 100ms, from asking for own cache space
Seeking time is ignored.
As shown in figs. 6-9, wherein Fig. 6 for EMD-ARIMA prediction model and only uses ARIMA prediction mould to simulation result
Type is to the prediction effect comparison diagram of mobile subscriber's request data, the as can be seen from the figure predicted value of EMD-ARIMA prediction model
It is very close with truthful data, and the prediction error of ARIMA prediction model is larger, it is seen that EMD-ARIMA prediction model is an advantage over
ARIMA prediction model is used alone.
Fig. 7 is that user requests the time delay of Web content with the changing rule of Zipf parameter alpha.When α takes smaller value, indicate
The request number of times of mobile subscriber is relatively uniform between each Web content, so the time delay difference that various cache policies provide
Less；With the increase of α value, user focuses more on the more several classes of request number of times to the request of Web content, due to prevalence
Spending the hybrid cache strategy mentioned in cache policy and the present invention is mainly that the more net of request number of times is stored in spatial cache
Network content, so the probability for obtaining Web content in each spatial cache increases, time delay reduces.And random cache strategy pair
Each Web content has storage, does not cache emphasis, so its probability drop for obtaining Web content in each spatial cache
Low, time delay is in rising trend.
Fig. 8 is that user requests the time delay of Web content with the changing rule of cell range.In this case, due to
Web content in family and node B cache space is constant, so the gain of user's request time delay mostlys come from field and delays
It deposits.When cell range is smaller, the distribution of user in the cell obtains Web content from the caching of field than comparatively dense, user
Probability it is bigger, so user request Web content time delay it is lower；With the increase of cell range, user is in the cell
Distribution it is gradually sparse, user from field cache in obtain Web content probability become smaller, user request time delay increase；When small
When area's range increases to a certain range, user obtains the probability of Web content close to 0, so user asks from the caching of field
Time delay is asked no longer to change substantially.
Fig. 9 is that user requests the time delay of Web content with the changing rule of buffer memory capacity.When the capacity of spatial cache is smaller
When, the probability that user obtains Web content from spatial cache is lower, and time delay is larger；With the increase of spatial cache capacity, use
The probability that family obtains Web content from spatial cache improves, and time delay reduces.
Generally speaking, at tri- kinds of Fig. 7~Fig. 9, the user that the hybrid cache strategy mentioned in the present invention provides is asked
Ask time delay all lower with the popularity cache policy of current mainstream than random cache strategy, it is seen that the mentioned strategy of the present invention is effective
's.
Although the illustrative specific embodiment of the present invention is described above, in order to the skill of the art
Art personnel understand the present invention, it should be apparent that the present invention is not limited to the ranges of specific embodiment, to the general of the art
For logical technical staff, if various change in the spirit and scope of the present invention that the attached claims limit and determine,
These variations are it will be apparent that all utilize the innovation and creation of present inventive concept in the column of protection.
Claims (1)
1. a kind of 5G forward pass mixture of networks edge cache low time delay method, which comprises the following steps:
(1), the 5G forward pass network edge buffer network scene between Core server, cell base station and user is constructed；
Cell base station obtains Web content by 5G forward pass network from Core server, then with communication is entire small
All mobile subscribers within the scope of area provide service, and every mobile subscriber provides service in radiation scope again for other users；
(2), the request that mobile subscriber issues out Web content, extract the requested Web content title of the mobile subscriber with
And corresponding request number of times；
Content name and corresponding request number of times are formed into information matrix again, and are stored in the database of cell base station；Its
In, every a line of information matrix represents a mobile subscriber to the request number of times of each Web content, and each column represent a network
Content name；
(3), the data in information matrix are pre-processed
Information matrix is numbered by row from 1 to N, then each column of information matrix are subjected to minimax normalized；
Wherein, x indicates request number of times of the mobile subscriber to some Web content, xminIndicate the smallest request number of times in the column,
xmaxIndicate maximum request number of times in the column；
(4), the data after pretreatment are clustered using density peaks clustering algorithm
(4.1), the distance between mobile subscriber is calculated
If each mobile subscriber makes requests M Web content, the mobile use of any two is calculated using Euclidean distance formula
Family usAnd utThe distance between dist (us,ut)；
Wherein, s, t ∈ [1, N], and s ≠ t,Indicate mobile subscriber usTo the request number of times of m-th of Web content；
(4.2), mobile subscriber u is calculatedsLocal density ρs；
ρs=∑ χ (dist (us,ut)-distcutoff)
Wherein, distcutoffIndicate truncation distance, function χ (x) meets:
(4.3), mobile subscriber u is calculatedsWith the higher user u of local densitytThe distance between δs；
As mobile subscriber usLocal density when being maximum value, then mobile subscriber usWith remaining mobile subscriber utThe distance between δs
Are as follows:
(4.4), it clusters
For all mobile subscribers, with local density ρsFor horizontal axis, distance δsFor the longitudinal axis, decision diagram is drawn in plane coordinates；
The point for being located at the decision diagram upper right corner is selected as cluster class center again, remaining each point is belonged to the cluster class nearest apart from it
In the affiliated class cluster in center, to clustering all points for L class, L={ l1,l2,…,lτ,…,lL, lτFor the movement in τ class
Number of users；
(5), the caching value of each Web content in each class cluster is calculated
The caching value of p-th of Web content in class cluster τ are as follows:
Wherein, fk,pFor the element value in information matrix, i.e. expression user ukTo the request number of times of Web content p；
After the completion of the caching value calculation of each Web content in each class cluster, it is at size by all caching value maintenances
The matrix Score of L*M；
(6), base station and hybrid cache strategy of each mobile subscriber when storing Web content are formulated
(6.1), node B cache strategy is formulated
Each of matrix Score element is subjected to descending arrangement, then Web content storage corresponding to preceding B element is arrived
In the spatial cache of base station, if Web content corresponding to some element duplicates, next network is stored in order
Content；
(6.2), mobile subscriber's cache policy is formulated
Each mobile subscriber caches the highest b Web content of respective request number of times in memory space；
(7), EMD-ARIMA prediction model is established
(7.1), mobile subscriber's request data x ' in all kinds of clusters obtained in step (3) is decomposed into using EMD decomposition algorithm more
A intrinsic modal components imfi(t), i indicates i-th of intrinsic modal components；
(7.2), to each imfi(t) ARIMA prediction model is constructed respectively；
Calculate each imfi(t) auto-correlation coefficientAnd PARCOR coefficients
Wherein, γ is intrinsic modal components imfi(t) length, αtFor imfi(t) sample value,For imfi(t) desired value；
Wherein,
Using delay period number as abscissa, auto-correlation coefficient is that ordinate draws autocorrelogram；Using delay period number as abscissa,
PARCOR coefficients are that ordinate draws partial autocorrelation figure；
Autocorrelogram is recycled to examine mobile subscriber to the stationarity of network content request data, it is right if data are unstable
Data carry out difference processing, and until data are steady, the number of accumulative difference processing is denoted as d, and the ginseng as ARIMA prediction model
Number；
Suitable ARIMA prediction mould is selected according to the truncation of time series autocorrelogram and partial autocorrelation figure and hangover property
Type, and ARIMA prediction model order is determined according to BIC criterion；
Wherein, n indicates each imfi(t) in component sample data number, c is constant, and y is unknown parameter number, and ε is white noise
Sequence,For the variance of ε；
ARIMA prediction model is constructed according to the parameter of ARIMA prediction model and order, recycles the ARIMA prediction model of construction
To imfi(t) component is predicted, and its predicted value is denoted as imfi(t)'；
(7.3), by all intrinsic modal components imfi(t) predicted value imfi(t) ' be added and asked as mobile subscriber in all kinds of clusters
Seek the predicted value of data；
(8), cache replacement policy is formulated
(8.1), using mobile subscriber in all kinds of clusters of EMD-ARIMA prediction model prediction to the request number of times of Web content, then
Calculate the caching value Score' of each request content in all kinds of clustersτ(p), it and in descending order arranges；
(8.2), not buffered Web content is judged whether there is, if there is not buffered Web content m', and its
Caching value Score'τ(p) it is higher than some and is stored in the Score of the Web content m in base station or mobile subscriber's spatial cacheτ
(p), then Web content m is then replaced with m' in advance in non-traffic peak period；
(9), mobile subscriber obtains Web content
(9.1), mobile subscriber issues some Web content and requests, mobile subscriber equipment terminal first can itself caching it is empty
Between in search whether to store the Web content in advance, if being stored with the Web content, be denoted as own cache, and directly from
This content is extracted in own cache space, the time used ignores；Otherwise, other into service range of user equipment terminal
User broadcasts the request to Web content, enters step (9.2)；
(9.2), it after other mobile subscribers receive request, searches whether to be stored with this content respectively from own cache space, such as
Fruit is stored with the content, then is denoted as neighborhood caching, and forward the Web content to request user, while recording time t used1；
Otherwise uncached to request user's reply, and enter step (9.3)；
(9.3), whether mobile subscriber sends to cell base station and requests, after cell base station receives request, search and deposit in node B cache
Requested Web content is contained, if so, being then denoted as node B cache, and submits the content to request user, while used in record
Time t2；Otherwise, the content is downloaded by 5G forward pass link in base station from Core server, is then forwarded to request user, together
Time t used in Shi Jilu3。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910515104.9A CN110418367B (en) | 2019-06-14 | 2019-06-14 | 5G forwarding network hybrid edge cache low-delay method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910515104.9A CN110418367B (en) | 2019-06-14 | 2019-06-14 | 5G forwarding network hybrid edge cache low-delay method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110418367A true CN110418367A (en) | 2019-11-05 |
| CN110418367B CN110418367B (en) | 2021-05-14 |
Family
ID=68359106
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910515104.9A Expired - Fee Related CN110418367B (en) | 2019-06-14 | 2019-06-14 | 5G forwarding network hybrid edge cache low-delay method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110418367B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111935781A (en) * | 2020-06-28 | 2020-11-13 | 中国科学院深圳先进技术研究院 | Control method of data sharing network, network system and related device |
| CN112020103A (en) * | 2020-08-06 | 2020-12-01 | 暨南大学 | Content cache deployment method in mobile edge cloud |
| CN112733060A (en) * | 2021-01-13 | 2021-04-30 | 中南大学 | Cache replacement method and device based on session clustering prediction and computer equipment |
| CN112839082A (en) * | 2020-12-31 | 2021-05-25 | 西安电子科技大学 | Heterogeneous edge cache allocation method, system, medium and application |
| CN113630742A (en) * | 2020-08-05 | 2021-11-09 | 北京航空航天大学 | Mobile edge cache replacement method adopting request rate and dynamic property of information source issued content |
| CN115884094A (en) * | 2023-03-02 | 2023-03-31 | 江西师范大学 | Multi-scene cooperation optimization caching method based on edge calculation |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020092026A1 (en) * | 2001-01-09 | 2002-07-11 | International Business Machines Corporation | Method and apparatus for broadcast delivery of content to a client-side cache based on user preferences |
| CN102291447A (en) * | 2011-08-05 | 2011-12-21 | 中国电信股份有限公司 | Content distribution network load scheduling method and system |
| WO2016023384A1 (en) * | 2014-08-12 | 2016-02-18 | 华为技术有限公司 | Cache content hitting method and communication system |
| CN106303704A (en) * | 2016-08-19 | 2017-01-04 | 上海交通大学 | A kind of DASH flow medium live system based on proxy server and method |
-
2019
- 2019-06-14 CN CN201910515104.9A patent/CN110418367B/en not_active Expired - Fee Related
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020092026A1 (en) * | 2001-01-09 | 2002-07-11 | International Business Machines Corporation | Method and apparatus for broadcast delivery of content to a client-side cache based on user preferences |
| CN102291447A (en) * | 2011-08-05 | 2011-12-21 | 中国电信股份有限公司 | Content distribution network load scheduling method and system |
| WO2016023384A1 (en) * | 2014-08-12 | 2016-02-18 | 华为技术有限公司 | Cache content hitting method and communication system |
| CN106303704A (en) * | 2016-08-19 | 2017-01-04 | 上海交通大学 | A kind of DASH flow medium live system based on proxy server and method |
Non-Patent Citations (2)
| Title |
|---|
| CHONGFU ZHANG1, MING NI, YAO ZHONG: "Density-Ratio Peak Based Semi-Supervised", 《IEEE ACCESS》 * |
| 孙毅，王鸿，宋荣方: "异构网络中两种边缘缓存策略的能耗分析", 《研究与开发》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111935781A (en) * | 2020-06-28 | 2020-11-13 | 中国科学院深圳先进技术研究院 | Control method of data sharing network, network system and related device |
| CN111935781B (en) * | 2020-06-28 | 2023-04-25 | 中国科学院深圳先进技术研究院 | Control method, network system and related device of data sharing network |
| CN113630742A (en) * | 2020-08-05 | 2021-11-09 | 北京航空航天大学 | Mobile edge cache replacement method adopting request rate and dynamic property of information source issued content |
| CN112020103A (en) * | 2020-08-06 | 2020-12-01 | 暨南大学 | Content cache deployment method in mobile edge cloud |
| CN112020103B (en) * | 2020-08-06 | 2023-08-08 | 暨南大学 | Content cache deployment method in mobile edge cloud |
| CN112839082A (en) * | 2020-12-31 | 2021-05-25 | 西安电子科技大学 | Heterogeneous edge cache allocation method, system, medium and application |
| CN112733060A (en) * | 2021-01-13 | 2021-04-30 | 中南大学 | Cache replacement method and device based on session clustering prediction and computer equipment |
| CN112733060B (en) * | 2021-01-13 | 2023-12-01 | 中南大学 | Cache replacement method and device based on session cluster prediction and computer equipment |
| CN115884094A (en) * | 2023-03-02 | 2023-03-31 | 江西师范大学 | Multi-scene cooperation optimization caching method based on edge calculation |
| CN115884094B (en) * | 2023-03-02 | 2023-05-23 | 江西师范大学 | Multi-scene cooperation optimization caching method based on edge calculation |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110418367B (en) | 2021-05-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110418367A (en) | A kind of 5G forward pass mixture of networks edge cache low time delay method | |
| Zhong et al. | A deep reinforcement learning-based framework for content caching | |
| CN106453495B (en) | A kind of information centre's network-caching method based on content popularit prediction | |
| CN108667653B (en) | Cluster-based cache configuration method and device in ultra-dense network | |
| CN112995950B (en) | Resource joint allocation method based on deep reinforcement learning in Internet of vehicles | |
| CN103455531B (en) | A kind of parallel index method supporting high dimensional data to have inquiry partially in real time | |
| CN108093435B (en) | Cellular downlink network energy efficiency optimization system and method based on cached popular content | |
| CN104426979A (en) | Distributed buffer scheduling system and method based on social relations | |
| CN106603646B (en) | A kind of information centre's network-caching method based on user interest preference | |
| Yang et al. | IoT information service composition driven by user requirement | |
| CN111314862B (en) | Caching method with recommendation under deep reinforcement learning in fog wireless access network | |
| CN110234155A (en) | A kind of super-intensive network insertion selection method based on improved TOPSIS | |
| CN113810931B (en) | Self-adaptive video caching method for mobile edge computing network | |
| CN107734482A (en) | The content distribution method unloaded based on D2D and business | |
| CN111491331A (en) | Network perception self-adaptive caching method based on transfer learning in fog computing network | |
| CN108521640B (en) | Content distribution method in cellular network | |
| CN109831790A (en) | Cooperation caching method based on brainstorming optimization algorithm in mist wireless access network | |
| CN116321307A (en) | Bidirectional cache placement method based on deep reinforcement learning in non-cellular network | |
| CN113472420B (en) | Satellite network cache placement method based on regional user interest perception | |
| Zhang et al. | DENA: An intelligent content discovery system used in named data networking | |
| CN108810139A (en) | A kind of wireless caching method based on Monte Carlo tree search auxiliary | |
| CN105407383A (en) | Multi-version video-on-demand streaming media server cluster resource prediction method | |
| CN108512765B (en) | Network content diffusion method based on network node distributed Pagerank | |
| CN112911614A (en) | Cooperative coding caching method based on dynamic request D2D network | |
| CN114245422A (en) | Edge active caching method based on intelligent sharing in cluster |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20210514 |